Disclosed are various approaches for automatically assigning weights to vertices of a skin or mesh that control how said vertices in the 3D model move under the influence of skeletal rotation and translation. A computing device can receive a first model weightings matrix. Next, the computing device can include adjusting the number of rows in the first model weightings matrix to generate an adjusted model weightings matrix with a number of rows that matches an input number of rows for a machine-learning model, each row in the adjusted model weightings matrix representing a vertex of a mesh applied to a three-dimensional model. Then, the computing device can apply the machine learning model to the adjusted model weightings matrix, to generate an output polygonal mesh model weightings matrix. Subsequently, the computing device can generate a second polygonal mesh model weightings matrix by adjusting the number of rows of the machine learning model output weightings matrix to match the number of rows of the initial polygonal mesh model weightings matrix.
Legal claims defining the scope of protection, as filed with the USPTO.
2. The system of claim 1, wherein the machine-readable instructions that adjust the number of rows in the weightings matrix to generate the adjusted model weightings matrix further cause the computing device to apply a nearest neighbor algorithm to the first model weightings matrix to generate the adjusted model weightings matrix.
3. The system of claim 1, wherein each value in the first model weightings matrix, the adjusted model weightings matrix, the output model weightings matrix, and the second model weightings matrix represents an intensity of a relationship between a vertex of a three-dimensional polygonal mesh and one or more skeletal bones or joints.
5. The system of claim 4, wherein the convolutional kernels of the convolutional neural network are one-dimensional convolutional kernels.
7. The method of claim 6, wherein adjusting the number of rows in the weightings matrix to generate the adjusted model weightings matrix further comprises applying a nearest neighbor algorithm to the first model weightings matrix to generate the adjusted model weightings matrix.
8. The method of claim 6, wherein each value in the first model weightings matrix, the adjusted model weightings matrix, the output model weightings matrix, and the second model weightings matrix represents an intensity of a relationship between a vertex of the three dimensional polygonal mesh and a bone or joint.
10. The method of claim 9, wherein the convolutional kernels of the convolutional neural network are one-dimensional convolutional kernels.
12. The non-transitory, computer-readable medium of claim 11, wherein the machine-readable instructions that adjust the number of rows in the weightings matrix to generate the adjusted model weightings matrix further cause the computing device to apply a nearest neighbor algorithm to the first model weightings matrix to generate the adjusted model weightings matrix.
13. The non-transitory, computer-readable medium of claim 11, wherein each value in the first model weightings matrix, the adjusted model weightings matrix, the output model weightings matrix, and the second model weightings matrix represents an intensity of a relationship between a vertex of a three-dimensional polygonal mesh and a bone or joint.
15. The non-transitory, computer-readable medium of claim 14, wherein the convolutional kernels of the convolutional neural network are one-dimensional convolutional kernels with variable sized filters.
Cooperative Patent Classification codes for this invention. Click any code to explore related patents in that topic.
June 16, 2022
May 28, 2024
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.